Chet Cooper: Tell us a bit more about Nao.
Olivier Joubert: Our project started in May 2009 when Aldebaran Robotics decided to adapt these humanoid robots for people with special needs. We had done research, mainly at the University of Notre Dame, the University of Connecticut, and more recently Vanderbilt University, to evaluate the interaction between Nao and children with autism. The results were promising. And because these institutions sought to develop new experiments and studies, they kept in close contact with Aldebaran, enabling us to get a very early understanding of how humanoid robots interact with children. After two years, we customized the robot to make it more useable for teachers working with children with autism.
Cooper: It’s my sense that working with children who are autistic was not your initial goal.
Joubert: That’s true. Bruno Maisonnier, the CEO of Aldebaran, felt strongly that robotics would be the technology of the 21st century. He was very passionate about robots, especially humanoid ones, because he felt the shape would make humans more comfortable interacting with them. So he created a company, worked on this product, and developed the first version of Nao in 2005. Then he began distributing it to different markets, including the research and university markets.
Our university partners started to explore using Nao with autistic children and the results were promising. And thanks to their research, our company realized we could do something meaningful in the field of special education. Two years ago we started to work actively on this project, and we now have a product that works well for teachers of children with autism.
Cooper: How much does a school have to pay for a Nao?
Joubert: For a robot loaded with a series of applications, plus an online interface that allows teachers to customize a teaching session, it’s $14,000. We call it the Autism Solution for Kids—or the ASK Nao solution. It includes the robot and a series of applications for children who have a large number of learning goals. We have an application that allows students to work on imitation, turn taking, emotion recognition and other important skills.
Let’s say you are a teacher planning what you would like to do that day with one of your students. You connect to an online interface that we provide, and then you input: “Today I want to work with Jordan.” On the online interface, you connect with Jordan’s “passport,” which gives you all the information related to him: his age, gender, learning goals, interests, the way he communicates, and other relevant information. From there, you decide what types of exercises you want to do with Jordan, and finally go to another interface and input: “I want to start this application with this child.” After you accept the customization for Jordan, you click Play Nao, and the robot starts to interact with Jordan.
The way Nao interacts with a child is inspired by the Applied Behavior Analysis (ABA) method, meaning that Nao will prompt the child to do something, asking her: Show me the picture of the crow, for instance, and then Nao will wait for the child to show the picture of the crow. If she picks the right picture, then Nao will reward her by doing a little dance. If it is a wrong answer, Nao will say, “Okay, that’s not the right one, we’ll try again.”
Cooper: Have you done an analysis of what it costs for Nao to work with a child vs. the cost of using a therapist?
Joubert: First I want to emphasize that we decided to use ABA because it’s the best known method and most used today, but we try to get inspiration from a variety of methods. Some of the applications are ABA-inspired, but others come from equally relevant sources. We get ideas from everywhere, but prefer to let researchers, who are experts in their field, do the testing in order to gain a better understanding of what’s possible with Nao. The relationship we have with researchers, and the beta tests we conduct in schools, show promising results. Nao is opening doors to the minds of children.
Cooper: What are your next steps?
Joubert: Today we are working with two different beta test schools: one is in the UK, and one is in New York. We want to adapt the solution not only for one or two children, but also for children worldwide. The data we collect is only one way to do research.
Cooper: Do you see the robot as a one-on-one interaction only? Or can it work with more than one child at a time?
Joubert: We have different applications. Some of them are more one on one, while others can be done in groups. For instance, when we are doing an ABA exercise, where Nao is prompting a child, waiting for an answer, and providing a reward, we see that as a triangle between Nao, the teacher, and the child. Nao will be the one to approach the child, while the teacher will be there to focus on the child, trying to get a better understanding of what the child needs, and also guide the her to successfully interact with Nao. In that case, it’s a one-to-one interaction, Nao plus the child plus the teacher all working together. There are also other applications, such as dance and collaborative games, which can be done with more than one child. In some beta testing, we have teachers who work with eight children and one robot at the same time.
Cooper: What other applications does Nao have for children?
Joubert: As a company, we decided to focus on autism because it’s what we know best. I have a PhD in neuropsychology, and I started to work with children with autism during a post-doc. What we have developed, so far, is something that can be used with other kinds of disabilities, as well. We might work with children who have some emotional or developmental challenges. We might even work with typical children to enhance memory. And we can also work on mentoring children on such academic skills as math, French or geography. So while the applications developed so far have been for children with autism, we can develop others to meet a range of needs. The possibilities are limited only by our imagination.
Cooper: I noticed that you have an online interface that allows one to track the progress of the work with a child.
Joubert: We decided to go with an online interface to connect teachers, parents and therapists who may not be in the same place. This allows all of them to access and share information about the child. This is all connected to Nao, so the robot is better able to “understand” if the child has provided the correct or incorrect response. Based on that information, it will track the performance and display it on a curve in the interface, allowing teachers to have a tool that is easy to use, program and/or customize.
Cooper: Must teachers, parents, and other caregivers use a password to get in to monitor these activities?
Joubert: Yes. When we send the robot to a school, users need login and password information, and someone in the school must be responsible for assigning icons for the teacher and parents in order to connect every child with the parents or guardians. Only people who are connected with the child can see the child’s information.
Cooper: Can one program the robot and download new curriculum remotely?
Joubert: Yes. We wanted to make it very easy for teachers to use. For instance, you could compare what we have done to the smart-phone design, where you can load the applications you need onto the device. Similarly, based on what you want to do with the children and the learning goals you want to work on, you will be able to go onto the platform and say, “I would like to use this video and this video.” You would just download them onto your robot. That way you are able to work on different aspects of the interaction with the children.
Some research shows that children are more attracted to interacting with a robot because of the technology and its predictability, and because the child will be able to respond faster or be less shy and more confident, which are skills they were already working on.
Cooper: You mentioned the interaction between children and the robot. Is it anecdotal that they interact with a robot in a different, more positive way? Has there been a large survey or study on this topic?
For instance, we’ve seen some evidence that children interact more with a robot than with a person or a passive toy. Since it is a robot, it is very predictable. The robot will always do the same thing the same way, and because it will not tire, a child can work with it much longer.
Cooper: Is there a camera? Do you videotape the children through the robot?
Joubert: Nao has two cameras: one to observe a child’s emotions, and the other to look at the child’s progress. At some point, we would like to create an application enabling Nao to record video and show the video to parents, but we’re not there yet.
Cooper: How much does the robot act like a robot? Can it go up a set of stairs? Can it lift things?
Joubert: Go on Google and you’ll be able to see Nao dancing, working and grasping objects. He’s also able to get a nice understanding of the environment, meaning that he’s able to recognize a sound and rotate his head towards the origin of the sound. He’s able to follow and recognize faces. These are things that Nao can do that are similar to a simplified human.
We had a teacher who worked with a small boy with autism, and when Nao entered the classroom, the child said, “I want Nao.” The teachers were very excited because it was the first time they saw this child speak. So this is a promising area. We’ve also seen children stand up and dance with Nao. It’s amazing how confident and relaxed they are. A researcher from the University of Connecticut developed an application where Nao plays a very specific rhythm on a drum, and then Nao interacts with the children while they play the instrument, and pretty soon the children can play the same rhythm as Nao.
Cooper: We were talking earlier about taking this technology international. How is that going in terms of using different languages and incorporating culture? For example, if you were doing this in Arabic?
Joubert: We have to take those things into consideration. Nao can speak, I think, about 13 or 14 languages, including English, French, German, Spanish, Italian, Portuguese and Japanese. He’s able to say the words and understand them. So on the device side, Nao can speak different languages, but we’ve developed an application, so far, to make Nao speak English because it’s the most-used language, and also Spanish, which is widely spoken in the US. It’s not that hard to add languages. We just had to start somewhere, so we started with English and Spanish.
Cooper: When you built the website and the applications for the different people to interface, did you think about accessibility being built into it as well?
Joubert: In terms of languages?
Cooper: No, meaning would people who are blind and who use the JAWS computer-screen reader program be able to navigate your interface?
Joubert: No, not yet. That is something we’ll ultimately do, though.
Cooper: Good. The percentages are such that most people don’t have disabilities, or at least obvious ones, but we always strive to build at a level that offers the highest accessibility possible, for anyone who comes into any field or occupation, whether it’s teaching, engineering or medicine, even. We know physicians who are blind.
Joubert: That’s very important feedback because we don’t want to design robots just for some people. We want to design robots for the betterment of humankind. We started with an interface aimed at teachers who don’t need accommodations, but our intention is that our interface will ultimately be accessible to everyone.
Cooper: In the US we have section 508, which means that if you’re a federal contractor, you have to make everything as accessible as possible, including your website and any computer programs, or the government won’t buy it from you. So there’s a push for this, but people still forget, especially in the programming stages, but I’m not trying to preach to you.
Joubert: No, this is the kind of feedback we need. Our products have been built on this kind of feedback, meeting with teachers, parents and journalists who say, “You should do that,” and we say, “Yes, we can do that.” We just have to do it, that’s all. Thanks a lot.
Cooper: Are you waiting to get more feedback before you expand? Do you have a goal in mind with the schools you’re working with?
Joubert: The main goal was to be sure the applications we have developed really match the needs of children with autism, and also to make the online interface very easy to use. Now that we’ve achieved these goals, we’re ready to enter the market. Right now, we’re only going with schools, therapy centers or autism clinics.
Cooper: Where do you manufacture the robots?
Joubert: In China, which is less expensive, but then we integrate most of the robot components in Paris. Aldebaran Robotics has developed everything, except for the programming language, which was subcontracted to developers. But Aldebaran developed 100 percent of the online interface, Nao and the system behind it, and most applications as well, except for the involvement of some outside developers. When I say outside developers, I mean people who like to program and who decided to help us develop more applications for children with special needs.
Maybe some teachers will be able to provide some specifications, and we will be able to implement them. One of the projects we have in mind, which is very important to us, is to design a platform where parents and teachers will be able to connect to something similar to a crowdfunding website and say, “I think your product is great and would be even better if it had an application that worked on this particular issue: I would like Nao to stand up, grab an object and give it to the child.” They will write a specification because they have the knowledge, and on the other side, we will have a community of developers who will connect to the same website and say, “Hey, I want to develop the application. I will make it because I want to help more people.”
Cooper: So the developer is similar to a model like WordPress, where people keep developing, and sometimes the apps are free and some are not?
Joubert: Yes, like an app on the smart phone. You have some people or company developing an application, and they will be able to make them available to everyone who uses a Nao.
Cooper: You didn’t partner with other companies that already have the technology you could build open source from? You did this from scratch?
Joubert: Nao is able to understand what you say through a text-to-speech algorithm, designed by one of our partners. So we do work with some other companies.
Cooper: What time frame are you looking at for entering the commercial arena, and is your distribution in place?
Joubert: We’re in 15 or 20 schools that bought the robot and now use it, even though the market is very new. We have entities in Western Europe and the US, who are available to meet with teachers in the schools and demonstrate how Nao works, and what they can do with it. We also have retailers working with us in the US and UK.
We spend a lot of time in schools to see how kids interact with Nao. We get a lot of feedback from teachers and parents saying how much they like it. Nao is kind of a teacher, too; it’s a way to teach children how to interact. When Nao is standing up by himself, the children stand up, too. When Nao sits down, the children sit down as well. It’s a new tool to make them understand some things. We want them to still be individuals, and yet be integrated into society. If you go on our website, we have testimonials from teachers, including blogs written by teachers who share their experiences with Nao. It’s amazing. It’s not just a commercial project; it’s more of a mission.
Thomas Chappell, MD: Chet, you’ve covered most of the things I was interested in, but I wanted to go back into a couple of areas and dig a little deeper. I wanted to ask about the data gathering or research aspect. As a medical doctor, we think of outcomes when we’re treating someone. Is someone following these kids and documenting their progress as they interact and learn with the robot?
Joubert: One of our partners has done some studies comparing the progress of kids with and without Nao. There are papers demonstrating the benefits of using robots with children with autism. In fact, Autism Speaks recently posted an article in which they talk about Nao and children with autism.
Chappell: As you get more of these robots on the market, will the price point start to come down? Do you have a lot of R&D costs that you’re trying to cover? Is the company on the stock market or planning to go public?
Joubert: As a company, we have a vision and an ethos: we want to design robots to better humankind. We want to develop useful robots that further what people can achieve. We also want to provide robots that, at some point, will be affordable to everyone, including schools and individuals. With each decision we make, we try to find the best balance between how much people can get from the product and the price of the product. We try to make it as affordable as possible and, as you say, more and more robots coming on the market will lower the price. The day will come very soon, when a robot is cheap enough for parents to buy. In terms of being traded on the stock market, we haven’t crossed that bridge yet.
In terms of R&D, much of that information is privileged, but we have 300 employees with the company, and most of them are engineers and researchers working on Nao—the software and hardware. We attract a lot of people who want to devote their lives to making
progress in this area. We also have about 20 patents now to cover some very specific aspects of our project.
Chappell: Where else do you think this technology might be applied?
Joubert: We have a special education project that would help institutions; it would be able to play with children; help them connect on Skype, check e-mail and coordinate with other personal devices. Robots can do a lot of different things; their potential is virtually unlimited.